Models you can keep
Save a trained model to your home folder. It is still there after the server restarts.
JupyterLab 4.6.3 Data
Free credit available Limited time
Jupyter TensorFlow and Keras TensorFlow imports in your first cell, with nothing toย install.
Selfhost.dev runs Jupyter TensorFlow as a zero-setup, one-click template: JupyterLab with TensorFlow and Keras preinstalled beside numpy, pandas and scikit-learn, on a dedicated server we provision, secure and run. It runs on the CPU; saved models stay in a home folder that persists.
Open a notebook and import tensorflow as tf. That is the whole setup.
Jupyter + TensorFlowone-click template ready
- installedTensorFlow 2.21.0, Keras
- data toolsnumpy, pandas, sklearn
- sign-inyour password, 12+
- runs onCPU
- kept on restartnotebooks, models
European 8 GB server$0.03/hr
In one line
What is Jupyter TensorFlow on Selfhost.dev?
What it is
Jupyter + TensorFlow is JupyterLab with TensorFlow and Keras added to numpy, pandas and scikit-learn. It is the template for deep learning notebooks that run on the CPU.
Where to start. Open the console with Jupyter + TensorFlow selected
Who it is for
Students learning Keras, developers prototyping a model and anyone tired of repairing a broken TensorFlow install.
When to use it
When a local TensorFlow install keeps failing or a training run should outlast your laptop's battery.
Server
Jupyter TensorFlow, with nothing to install
The usual route to TensorFlow in Jupyter is a virtual environment, pip install tensorflow, then ipykernel to register that environment as a kernel. Miss one step and the notebook says No module named tensorflow. Here TensorFlow is installed in the kernel you open.
The stack, end to end
Callers
Your browser
Edge
HTTPS
Your project server
Container
JupyterLab
Kernel
Python
Library
TensorFlow + Keras
Kept
Home folder
To sign in
Your password
| TensorFlow in Jupyter by hand | Jupyter TensorFlow on Selfhost.dev |
|---|---|
| Create and activate a virtual environment | One environment, already set up |
| pip install tensorflow into it | TensorFlow installed |
| Register it with ipykernel install | The kernel you open already has it |
| Restart the kernel after !pip install | No restart before the first import |
Only your password opens JupyterLab and its TensorFlow kernel. Deploy the stack
Container
Keras in Jupyter, beside TensorFlow
Keras is TensorFlow's high-level API, so it arrives in the same kernel. Define a model with keras.Sequential, compile it and call fit, then save the trained model in your home folder.
pandas for the data
Load and shape a dataset with pandas, then pass the arrays to Keras in the same notebook.
scikit-learn for a baseline
Split the data, scale features and score a simple model before you train a network.
What you need
Jupyter TensorFlow requirements
Jupyter TensorFlow starts on an 8 GB server, the size the console picks for it, on x86 or ARM. Project servers have no GPU, so TensorFlow runs every model on the CPU. That suits learning, prototypes and small models.
What it needs
- Server
- 8 GB, chosen by the console; x86 or ARM
- Processor
- CPU only: Keras models, prototypes, inference
- Password
- 12 characters or more, set when you deploy
- Domain
- Optional; the generated HTTPS address works first
Server size: 8 GB, enough for Jupyter + TensorFlow. Deploy on an 8 GB server
Quickstart
How to start Jupyter TensorFlow in 3 steps
Launch, sign in, import.
-
Launch the template
Launch opens Jupyter + TensorFlow. Pick a password of 12 characters or more and accept the no-mining terms. Before a first top-up, a platform admin approves it. The template deploys in about twenty minutes.
-
Sign in
Go to the HTTPS address in the console and enter your password in the Password or token box. No username is asked for.
-
Import TensorFlow
Run import tensorflow as tf in a new Python notebook. Keras, numpy, pandas and scikit-learn are installed beside it.
Deploy Jupyter + TensorFlowStep 1 starts in the console.
Launch Jupyter + TensorFlow Billing
Jupyter TensorFlow cost
Jupyter TensorFlow is billed as one server by the hour from prepaid credits, with no charge per training run. A European 8 GB server costs $0.03/hr, roughly $23 a month.
$0.03
/hr, about $23 a month at 730 hours
European 8 GB server, $0.03/hr Deploy Jupyter + TensorFlow | Item | Cost |
|---|---|
| The Jupyter + TensorFlow template | $0 |
| Bandwidth to your notebooks | $0, included in the server price |
| Optional daily backups | +20% of the server price |
| Models you train and save | $0 |
| Server | RAM, cores, disk | Where | Rate | About a month |
|---|---|---|---|---|
cx33 (recommended) | 8 GB, 4 cores, 80 GB | Europe | $0.03/hr | $23.36 |
cax21 (ARM) | 8 GB, 4 cores, 80 GB | Europe | $0.04/hr | $29.20 |
cx43 (room to grow) | 16 GB, 8 cores, 160 GB | Europe | $0.06/hr | $43.22 |
ccx13 | 8 GB, 2 cores, 80 GB | US | $0.16/hr | $119.28 |
cpx32 | 8 GB, 4 cores, 160 GB | Singapore | $0.19/hr | $135.63 |
Monthly figures assume 730 hours. The US and Singapore servers need a first top-up.
If your balance reaches zero
With a card on file or a past top-up, Jupyter + TensorFlow runs for 5 more days after your balance reaches zero, then stops. Top up within 24 hours to restart it; after that it is deleted, with a 14-day copy you can restore. Welcome-credit-only servers are deleted at zero.
Price a 16 GB server for bigger training sets in the calculator, no signup.
Options
Jupyter TensorFlow vs the TensorFlow Docker image
TensorFlow's Docker guide runs tensorflow/tensorflow:latest-jupyter with -p 8888:8888, then has you copy a 127.0.0.1 URL from the terminal and bind-mount a folder so files survive. Jupyter TensorFlow on Selfhost.dev is that container already running, on a server of its own.
| TensorFlow Docker image | Jupyter TensorFlow on Selfhost.dev | |
|---|---|---|
| Start | docker run with -p 8888:8888 | Already running |
| Sign-in | A token URL copied from the terminal | Your own password |
| Your files | Lost unless you bind-mount a folder | Kept in your home folder |
| Address | 127.0.0.1 on your machine | HTTPS, from any browser |
Status
Five checks before your first import
- Every service is health-checked every 60 seconds.
- A stalled deploy is retried automatically, without you starting over.
- Done means the URL answers, not just that a container started.
- Locked from the first request. The password you set at deploy is the only way in.
- You see the release it runs. The console flags Update available when a newer one ships.
~20 min to deploy
The console estimates about twenty minutes for the Jupyter + TensorFlow template after the server is ready; first-time server provisioning adds time.
Notes
Jupyter TensorFlow limitations
- Built for the CPU. Project servers have no GPU. The template suits learning Keras, prototypes, small networks and predictions from saved models.
- One workspace per password. Anyone holding the password shares the notebooks and kernels. A person who needs privacy gets a deployment of their own.
- A larger image, a slower first deploy. TensorFlow adds to the scientific Python stack, so the console allows about twenty minutes.
Nothing here in your way? The template deploys in about twenty minutes.
Launch Jupyter + TensorFlow Planning
Training models for a course or a team?
Tell us the dataset size and how many notebooks run at once. We reply with the server we would choose and its monthly price.
Help
Frequently asked questions
Can I use TensorFlow in a Jupyter Notebook?
Yes. TensorFlow runs in any Jupyter notebook whose kernel has it installed, which is where most setups break. On this template it is already installed in the Python kernel, so import tensorflow as tf works in the first cell with no venv, pip or ipykernel step.
Does Jupyter TensorFlow use a GPU?
No. Project servers have no GPU, so this template runs TensorFlow on the CPU and every model trains on the processor. That covers learning Keras, prototypes, small networks and predictions from saved models. Size the server for the data and models you load; a 16 GB server gives more room.
Is Keras included with Jupyter TensorFlow?
Yes. Keras is installed with TensorFlow, so you can build a model with keras.Sequential or the functional API in the same notebook. numpy, pandas and scikit-learn sit beside it for loading data, splitting it and scoring a baseline before you train a network.
Do my trained models stay after a restart?
Yes. Save a model into your home folder and it is still there after the server restarts, along with your notebooks and packages added with pip install --user. Keep a second copy of models you cannot retrain quickly, in a repository or a download.
How is Jupyter + TensorFlow different from JupyterLab?
It is the same JupyterLab with TensorFlow and Keras added to the scientific Python stack, so the image is larger and the console allows about twenty minutes. Without deep learning, JupyterLab hosting is the lighter template. For PyTorch, Jupyter + PyTorch brings torchvision and torchaudio.
Why might a new Jupyter + TensorFlow wait for approval?
Before a first top-up, each new Jupyter + TensorFlow waits for a platform admin to approve it. Its name is held for you meanwhile; the decision arrives by email. Organisations that have topped up skip the wait. Crypto mining is banned on every server. Delete the service to withdraw a request.
TensorFlow and Keras, installed. Your first model starts now.
- TensorFlow and Keras in the Python kernel, ready to import
- Saved models kept in your home folder
- A European 8 GB server, about $23 a month
About Selfhost.dev
Selfhost.dev is a managed self hosting platform for developers and small teams. Open source apps and databases on dedicated servers you control, billed by the hour, in six locations across three continents.
Best for: Deep learning notebooks beside the apps and data they use, a vector database for embeddings and managed PostgreSQL, MySQL and Redis.